BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sensible Priors for Sparse Bayesian Learning - Oliver Stegle (Univ
 ersity of Cambridge)
DTSTART:20071008T101500Z
DTEND:20071008T111500Z
UID:TALK8046@talks.cam.ac.uk
CONTACT:Philip Sterne
DESCRIPTION:Journal Club on : "Sensible Priors for Sparse Bayesian Learnin
 g".\n\nJoaquin Quiñonero Candela\; Edward Snelson come to present this an
 d for discussion.\n\nSparse Bayesian learning suffers from impractical\, o
 verconfident predictions where the uncertainty tends to be maximal around 
 the observations. We propose an alternative treatment that breaks the rigi
 dity of the implied prior through decorrelation\, and consequently gives r
 easonable and intuitive error bars. The attractive computational efficienc
 y is retained\; learning leads to sparse solutions. An interesting by-prod
 uct is the ability to model non-stationarity and input-dependent noise.\n\
 n------\nhttp://research.microsoft.com/research/pubs/view.aspx?tr_id=1364
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR